I had a random thought while filing bug 1427313. We're currently tracking build times and sccache hit rates, but a few rust crates count for 1 hit or miss each, but take a very long time to build when we get a miss (I'm looking at you, style). Which means there are many cases where we see a 0.99 or 1.0 hit rate, but bad build times. While build time gives some insight to that, one always needs to look at sccache logs (possibly lots of them) to figure out whether sccache is involved in build times regressions, or whether it's something else. So here's my thought: maybe the right metric to track with perfherder is the cumulative time spent invoking compilers by sccache.

This sounds useful. And it is related to metrics. So chaining to fastci tracker.